Molecular cytogenetics in haematological malignancy: current technology and future prospects
- First Online:
- Cite this article as:
- Kearney, L. & Horsley, S.W. Chromosoma (2005) 114: 286. doi:10.1007/s00412-005-0002-z
- 233 Views
Cytogenetics has played a pivotal role in haematological malignancy, both as an aid to diagnosis and in identifying recurrent chromosomal rearrangements, an essential prerequisite to identifying genes involved in leukaemia and lymphoma pathogenesis. In the late 1980s, a series of technologies based around fluorescence in situ hybridisation (FISH) revolutionised the field. Interphase FISH, multiplex-FISH (M-FISH, SKY) and comparative genomic hybridisation (CGH) have emerged as the most significant of these. More recently, microarray technologies have come to prominence. In the acute leukaemias, the finding of characteristic gene expression signatures corresponding to biological subgroups has heralded gene expression profiling as a possible future alternative to current cytogenetic and morphological methods for diagnosis. In the lymphomas, high-resolution array CGH has successfully identified new regions of deletion and amplification, providing the prospect of disease-specific arrays.
Refinements in cytogenetic techniques over the past 30 years have allowed the increasingly sensitive detection of chromosome abnormalities in haematological malignancy. In particular, the advent of fluorescence in situ hybridisation (FISH) has provided significant advances in both diagnosis and research of leukaemias and lymphomas. This review will concentrate on the most exciting recent developments using this wide-ranging technology and also provide speculation into the future roles for molecular cytogenetics in this rapidly changing field.
One of the most significant advances in diagnostic leukaemia cytogenetics has been the use of interphase FISH. The widespread availability of good quality commercial probes for the characteristic leukaemia-associated fusion genes has been instrumental in this success. More recently, “split-apart” probes for specific leukaemia-associated gene rearrangements have come to prominence (van der Burg et al. 2004; van Zutven et al. 2004). In this case, the two differentially labelled probes are normally located next to one another on the chromosome and only move as the result of a rearrangement. The value of this design, compared to gene fusion FISH, is the lower rate of false positives. Chance “splitting apart” is rare, compared to chance co-localisation using gene fusion probes. It is also independent of the partner gene, making it highly suitable for promiscuous genes such as MLL, TEL (ETV6) or NUP98. Split-apart probes are now becoming available for all of the clinically relevant chromosome rearrangements in acute lymphoblastic leukaemia (ALL) (reviewed in van der Burg et al. 2004).
Karyotype is an important prognostic indicator in childhood ALL with the presence of the Philadelphia translocation (resulting in the BCR–ABL gene fusion), MLL rearrangements and near-haploidy associated with a poor outcome, and high hyperdiploidy and the TEL–AML1 gene fusion associated with a favourable prognosis (reviewed in Harrison 2001). However, the incidence of normal karyotypes or cytogenetic failures is approximately 20%, due to a variety of factors including poor morphology of ALL chromosomes and the propensity of ALL blast cells to apoptose in culture. The Leukaemia Research Fund/United Kingdom Cancer Cytogenetics Group Karyotype Database in Acute Leukaemia (Harrison et al. 2001) has reported the results of an intensive interphase FISH screening program of over 2,000 patients entered onto the MRC ALL'97 treatment trial using probes for BCR–ABL, TEL–AML1 and MLL rearrangements (Harrison et al. 2005). Overall, this increased the success rate to 91% and the abnormality detection rate to 89%. Importantly, a significant number of additional BCR–ABL fusion positive cases and MLL rearrangements were identified by this procedure. In addition, interphase FISH with selected centromeric probes identified hidden hyperdiploidy in 33% of cases with a failed karyotype and in 59% with a normal karyotype (Harrison et al. 2005). It is clear that this strategy has been very effective in ensuring that patients are assigned to the appropriate risk groups.
The technique of combined immunophenotyping and FISH (also called FICTION) has provided a powerful tool for many research applications. One elegant application for this has been to investigate the natural history of pediatric leukaemia, helping to show that specific leukaemia-associated translocations are present in a subset of normal cord progenitor cells (Mori et al. 2000, reviewed in Greaves and Wiemels 2003). More recently, Maia et al. (2004) used this approach to identify specific trisomies in the CD34+CD19+ B cell lineage precursor cells of stored cord blood from an individual who later developed hyperdiploid ALL. This supports the idea that hyperdiploid ALL arises in a B-cell progenitor and that hyperdiploidy is an early, prenatal event in the pathogenesis of this common childhood ALL subtype. Another recent application for FICTION has been the investigation of lymphoma-associated endothelial cells for the presence of specific lymphoma-associated translocations (Streubel et al. (2004). Surprisingly, both primary and secondary lymphoma-associated chromosome abnormalities were found in a high percentage of the microvascular endothelial cells of B-cell lymphomas. This intriguing finding may indicate that the lymphoma arises in a multipotent progenitor cell capable of differentiation into haemopoietic and endothelial cells. Other possibilities include cell fusion between the lymphoma cells and microvascular cells, or gene transfer by apoptosis of tumour cells. Nevertheless, the phenomenon warrants further investigation.
Multiplex-FISH (M-FISH, SKY, COBRA-FISH)
The evolution of chromosome painting techniques to encompass 24-colour, FISH-based karyotyping (M-FISH, SKY, COBRA-FISH) has been one of the great successes of the past decade (Speicher et al. 1996; Schrock et al. 1996; Tanke et al. 1999). This technology has progressed even further with chromosome arm-specific, region-specific and subtelomeric probe sets (reviewed in Langer et al. 2004). However, hopes for this technology to discover new, clinically significant chromosome rearrangements in acute myeloid leukaemia (AML) patients with a normal karyotype have not been realised. Although early M-FISH and SKY studies did uncover cryptic clones and cryptic versions of known leukaemia-associated rearrangements, no new recurrent translocations were identified in normal karyotype AMLs using this whole chromosome painting based approach (Zhang et al. 2000; Mohr et al. 2000). Our studies (using multicolour subtelomeric probe FISH) did identify a new recurrent translocation, t(5;11)(q35;p15.5), in 7% of childhood AML cases with a normal karyotype (Brown et al. 2002). However, it appears that other mechanisms of leukaemogenesis (not evident as gross chromosomal changes) are important in this AML subgroup. This is borne out by gene expression profiling studies (Debernardi et al. 2003; Bullinger et al. 2004). The real value of 24-colour FISH lies in unravelling complex karyotypes. Mrözek et al. (2002) used SKY to study 29 adult AML patients with complex karyotypes, with at least one abnormality not resolvable by G banding. In this study, a high proportion of apparent deletions were shown to be cryptic translocations. Overrepresentation of chromosomes 11 and 21 was found in a high proportion of patients, and confirmed to involve MLL and RUNX1 amplification. Overall, karyotypes described as complex were found to be even more complex. These findings are typical of other studies of complex karyotypes in AML and MDS (Mathew et al. 2001; Schoch et al. 2002; Van Limbergen et al. 2002).
In childhood ALL approximately 80% of karyotypes are abnormal, with the remainder due to failed cytogenetics or an apparently normal karyotype. In some ALL subgroups, the leukaemic blasts have an increased propensity to apoptose, and a normal karyotype may reflect this. Interphase FISH screening can increase the abnormality rate significantly (Harrison et al. 2005). Because of the inherent poor chromosome morphology in ALL, SKY and M-FISH may also provide a valuable adjunct to karyotyping, particularly in cases of high hyperdiploidy and complex karyotypes (Elghezal et al. 2001; Lu et al. 2002). In T-ALL, a normal karyotype is reported in 30–40% of patients. Helias et al. (2002) used a variant of M-FISH (IPM-FISH) on patients with complex karyotypes in a range of haematological disorders. This identified three T-ALL cases (reported as normal by RHG banding) with the same cryptic translocation, t(5;14)(q35;q32). All were children with the same blast cell immunophenotype and clinical features. More recently, it has been demonstrated that this rearrangement is associated with overexpression of the homeobox gene HOX11L2, and defines a subgroup of T-ALLs with a poor prognosis (Balleriniet al. 2002). This is therefore a significant new and recurring translocation in a high percentage of childhood T-ALL cases revealed by the new technology. The recent finding of another cryptic rearrangement, inv(7)(p15q34), in a subset of T-ALLs with aberrant HOX11 expression indicates that this may be a common mechanism in T-ALL (Speleman et al. 2005).
Insights into high hyperdiploidy in childhood ALL
High hyperdiploidy, defined as >50 chromosomes, occurs in 30% of childhood leukaemias and constitutes a distinct subgroup characterised by a favourable prognosis. It is of interest that there is a non-random pattern of chromosomal gains, with chromosomes X, 4, 6, 10, 14, 17, 18 and 21 usually present, and an association of some trisomies with a specific prognosis (Moorman et al. 2003). However, while the pathogenetic consequences of chromosomal translocations on leukaemia are beginning to emerge, the transforming effects of additional chromosomes on leukaemia pathogenesis are unknown. There is evidence that the additional chromosomes arise prenatally (Panzer-Grumayer et al. 2002; Maia et al. 2003, 2004) and as the consequence of a single mitotic event (Onodera et al. 1992). It has also been suggested that imprinting effects, related to the parental origin of the duplicated chromosomes, might be responsible. Paulsson et al. (2003, 2004) have investigated the possible mechanisms of formation and the parental origin of the trisomic chromosomes in hyperdiploid ALL using microsatellite PCR for polymorphic microsatellite markers on all human chromosomes. By interpreting the allelic ratios on disomic and tetrasomic chromosomes, they concluded that there was no preferential gain of either maternal or paternal chromosomes in hyperdiploid ALL, and no evidence for uniparental disomy (UPD) or imprinting effects. Their results support the simultaneous gain of chromosomes, rather than polyploidisation and loss, as a mechanism for formation of a hyperdiploid karyotype. In another study, Haas et al. (2004) investigated the acquisition of X chromosomes in hyperdiploid ALL using a dual-colour FISH assay for X-inactivation-specific transcript (XIST) RNA and X centromere DNA. Two out of three X chromosomes in the leukaemic samples (both male and female patients) were active, confirming that the active X is duplicated in cases with acquired trisomy X. This confirms concept of hyperdiploidy arising as a consequence of a single non disjunction event.
Our studies of high hyperdiploid ALL (Gruszka-Westwood et al. 2004) used the chromosome-based global gene expression profiling method comparative expressed sequence hybridisation (CESH) (Lu et al. 2001) to examine relative expression. Overall, regions of over-expression by CESH corresponded to the presence of trisomies, with some peaks of expression observed along the length of the trisomic chromosomes. Tetrasomy resulted in higher expression and more even profiles. Increased expression without underlying trisomy occurred at a small number of regions in a proportion of cases. However, the use of different normal reference cell types indicated that at least some of these regions reflected cell type-specific expression. Gene expression profiling studies of hyperdiploid ALL using Affymetrix arrays have confirmed our CESH study (Yeoh et al. 2002; Ross et al. 2003). In these studies, the majority of class discriminating genes for high hyperdiploid ALLs were located on chromosomes X and 21, the chromosomes most often tri- or tetrasomic. In addition, the class discriminating genes had average fold increase of 2, consistent with a gene dosage effect for the commonly trisomic chromosomes.
Comparative genomic hybridisation (CGH, array CGH)
Comparative genomic hybridisation (CGH) is a method for determining copy number gains and losses between two samples of DNA, by competitively hybridising differentially labelled DNA to metaphase chromosomes (Kallioniemi et al. 1992). CGH has been a valuable tool in the analysis of lymphomas, as knowledge of the genomic alterations in this heterogeneous group of diseases is important for the classification, prediction of outcome and an improved understanding of the biology (reviewed in Lichter et al. 2000). Metaphase CGH has revealed high level of amplifications in 10% of non-Hodgkin's lymphomas, providing a starting point for the identification of genes with a role in pathogenesis (Lichter et al. 2000). CGH analysis of sequential biopsies has been used to identify copy number alterations associated with transformation of follicular lymphoma to diffuse large B-cell lymphoma (Hough et al. 2001;Viardot et al. 2001). More recently, microarray-based formats (array CGH, also called matrix-CGH) using large insert genomic clones, cDNAs or oligonucleotides have replaced metaphase chromosomes (reviewed in Albertson and Pinkel 2003). Array CGH provides the advantage over metaphases of a higher resolution, and the ability to directly map the copy number changes to the genome sequence. The first arrays of this type contained clones spaced at approximately 1 Mbp across the genome (Fiegler et al. 2003; Greshock et al. 2004). However, high-resolution tiling path arrays, consisting of overlapping BAC clones, are now available, increasing the resolution of this approach even further (Kryzwinski et al. 2004; Ishkanian et al. 2004; de Leeuw et al. 2004).
In the lymphomas, the improved resolution of array CGH formats has increased the number of genomic aberrations identified, and a picture is beginning to emerge of the clinical associations. Analysis of aggressive B-cell NHLs using array CGH with a dedicated B-cell lymphoma chip revealed cryptic amplifications involving (among others) BCL2, REL, CCND1, CCND2, JAK2, FGF4 and MDM2 (Wessendorf 2003). Importantly, a proportion of amplifications detected by array CGH were undetected by metaphase CGH. Martinez-Climent et al. (2003) found novel regions of genomic copy number change accompanying the transformation of follicular lymphoma to diffuse large B-cell lymphoma. In combination with expression array analysis, this provided a more comprehensive picture of the transformation process than either technique alone.
In B-CLL, the pioneering interphase FISH studies of Döhner et al. (2000) were the first to uncover the association between known genomic aberrations (deletions and trisomies) and survival in B-CLL. Deletions of 17p and 11q in particular were associated with a poor outcome. Schwaenen et al. (2004) described the development of a B-CLL chip containing genomic regions recurrently imbalanced in CLL, genes with a putative role in B-NHL, as well as known proto-oncogenes and tumour suppressor genes. Validation of this array, in comparison with interphase FISH, was carried out on 106 patients with B-CLL. This showed a high specificity and sensitivity compared to interphase FISH, and also identified two new regions of genomic imbalance including trisomy 19 and gain of MYCN. An important caveat is the requirement for >50% of abnormal cells. For 33–53% cells, array CGH detected lower percentages of trisomies and deletions than interphase FISH.
An updated version of the B-CLL chip has been used for array CGH of mantle cell lymphoma (Kohlhammer et al. 2004). In this study, array CGH identified the same spectrum of abnormalities as metaphase CGH. However, a 50% higher number of genomic aberrations was found using array CGH. For several regions, consensus critical regions of 400 kb–15 Mbp were defined and candidate genes identified. For example, an 8p minimal deleted region of 2.4 Mbp was defined, containing a number of candidate genes including several of the family of TRAIL receptors. The delineation of critical regions in this way holds promise for the identification of target genes and future new therapeutic options. Future clinical trials incorporating such strategies (alongside expression arrays) are set to evaluate their importance in risk assessment.
Array CGH of the acute leukaemias has lagged behind that of the lymphomas, although reports are now beginning to appear at scientific meetings. The majority of acute leukaemias are well characterised by recurrent (usually balanced) chromosomal rearrangements. However, a number of different leukaemia subgroups are ripe for investigation with array CGH, particularly those with no clinically relevant cytogenetics. In AML, the high percentage of apparently normal karyotypes represents one such group. Walter (2004) reported cryptic deletions and gains in 16% of AML patients with normal cytogenetics using a 1-Mbp resolution BAC array. In childhood ALL, the “normal” and “other” cytogenetics groups (for which there is no prognostic information) have been investigated using array CGH (C. Harrison and J. Strefford, personal communication). This revealed submicroscopic deletions in 50% of cases, some involving single clones. Overall, the abnormality rate increased from 86% by G banding to 100% using this approach. An additional advantage of array CGH is to define the extent of deletions and amplifications. Strefford et al. (2004) used array CGH to characterise the amplified region containing AML1 in the poor prognosis subgroup of childhood ALLs with duplicated chromosomes 21. A combination of array CGH, FISH and expression arrays identified a common deleted region of 10 Mbp, and 60 overexpressed genes, of which six were localised to the 21q amplicon. One of the most stunning applications for array CGH was demonstrated in the study by Graux et al. (2004) to fully elucidate the nature of the NUP214-ABL amplicon in T-ALL. They used a 1-Mbp BAC array to define the amplified region on 9q34 and to show that this was accompanied by deletion of the tumour suppressor genes CDKN2A and CDKN2B. This illustrates the power of this combined approach in unraveling the pathogenesis of this poor prognosis type of leukaemia.
Gene expression profiling in acute myeloid leukaemia (AML)
Currently, cytogenetic analysis (incorporating FISH) provides important information for classification and treatment stratification of newly diagnosed AML patients. However, cytogenetic analysis provides no clues in the case of AML with a normal karyotype, which accounts for 30–40% of cases. This group, along with various numerical and structural abnormalities, is assigned an intermediate risk, and it is this group which presents the biggest challenge for treatment options. Gene expression profiling studies have clearly shown that the characteristic chromosome rearrangements in AML are associated with distinct expression signatures (Debernardi et al. 2003). To date, the largest study to use gene expression profiling found only 16 different distinct profiles (clusters), indicating that AML may not be as heterogeneous as previously thought, and that there may be just a few common pathways to the development of leukaemia (Valk et al. 2004). Some of these corresponded to the well known cytogenetic subgroups, whereas others (notably rearrangements involving MLL) were not restricted to a single cluster. In another study, the potential of gene expression profiling to provide prognostic information was tested. Bullinger et al. (2004) used cDNA arrays and unsupervised hierarchical clustering to identify different classes of leukaemia. Gene expression signatures characterised known cytogenetic groups, and identified new subtypes. AML cases with a normal karyotype could be subdivided into two classes (Groups I and II). FLT3 mutations were more common in Group I (which had also decreased survival). Similarly, the inv(16) and t(8;21) groups could also be subdivided. Secondly, they applied both supervised and unsupervised strategies to develop gene expression-based outcome predictors. This identified a class predictor of 133 genes, which was successful in defining good outcome and poor outcome subgroups. One of the aims of this type of analysis is to develop a diagnostic tool to use expression profiling to accurately predict all clinically relevant subtype of leukaemia, to replace the current combination of disciplines, requiring morphology, immunophenotyping and cytogenetics. There is some evidence (from a recent meeting abstract) that this will be achievable (Haferlach et al. 2004).
Single nucleotide polymorphism arrays reveal a new mechanism in AML pathogenesis
While CGH can give information on copy number, it provides no information on the parental origin of chromosomes. In some cancer-associated genetic diseases, a single parental origin, uniparental disomy (UPD), has been associated with imprinted genes in the homozygous regions (reviewed in Jirtle 1999). Oligonucleotide arrays allowing genotyping of thousands of single nucleotide polymorphisms (SNPs), the most abundant form of genetic variation in the human genome, have been applied to the analysis of loss of heterozygosity (LOH) in paired normal and tumour samples (Bignell et al. 2004; Matsuzaki et al. 2004). The Affymetrix 10K SNP chip allows the visualisation of SNPs and LOH data along the length of chromosomes and also an assessment of copy number (by fluorescence intensity). Raghavan et al. (2005) used the Affymetrix 10K SNP array to investigate AMLs with either a normal karyotype or non-recurrent chromosome abnormalities, to determine the (so far obscure) pathogenesis in this group. Surprisingly, they found large regions of homozygosity in 20% of normal karyotype AML cases that could not be accounted for by chromosomal loss or gain (by FISH). One explanation for this observation is somatic recombination resulting in large regions of UPD. The possible effect of this “acquired UPD” could be to unmask mutations of leukaemia-associated genes. Indeed, this appears to be the case in one patient who was homozygous for a CEBPA mutation. The regions involved in this phenomenon were large, and appeared to be non-random. Therefore, this appears to be a major new mechanism of leukaemogenesis in AML.
Conclusions and future prospects
Molecular cytogenetics now has an established role in leukaemia and lymphoma diagnosis, as well as providing tools for research. Almost 20 years since its first description, FISH still has the ability to surprise, illustrated by the recent discovery of the NUP214–ABL1 fusion gene identifying a new subgroup of childhood ALL. CGH has been a very powerful technique in lymphoma diagnosis and research, and with the advent of high-resolution array CGH, promises to continue to stretch cytogenetic analysis by defining new genomic regions involved in disease pathogenesis. It is clear from recent expression and oligonucleotide arrays that each of these techniques offers a unique insight into subtypes of leukaemia that are currently not well understood. An important caveat here is the urgent need to improve bioinformatics approaches, due to the rapid generation of an enormous amount of data and the current lack of appropriate data processing tools and standardization. Nevertheless, it is likely that the future will employ a combination of cytogenetics, expression, and oligonucleotide arrays for a full diagnostic picture and that in the process, our understanding of the pathogenesis of haematological malignancies will be enhanced.
We wish to acknowledge the Medical Research Council UK and the Leukaemia Research Fund UK for financial support. We also thank Alan Mackay and the Breakthrough Breast Cancer microarray facility, Institute of Cancer Research for access to and advice on the use of the 1-Mbp BAC arrays.